[PATCH 3/3] mm/vmalloc: Cache the vmalloc memory info

[Ingo Molnar <mingo@kernel.org>]

Linus reported that glibc (rather stupidly) reads /proc/meminfo
for every sysinfo() call, which causes the Git build to use
a surprising amount of CPU time, mostly due to the overhead
of get_vmalloc_info() - which walks a long list to do its
statistics.

Modify Linus's jiffies based patch to use the newly introduced
vmap_info_changed flag instead: when we cache the vmalloc-info,
we clear the flag. If the flag gets re-set then we'll calculate
the information again.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: linux-mm@kvack.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 mm/vmalloc.c | 22 ++++++++++++++++++++--
 1 file changed, 20 insertions(+), 2 deletions(-)

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index d21febaa557a..ef48e557df5a 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -2702,7 +2702,7 @@ static int __init proc_vmalloc_init(void)
 }
 module_init(proc_vmalloc_init);
 
-void get_vmalloc_info(struct vmalloc_info *vmi)
+static void calc_vmalloc_info(struct vmalloc_info *vmi)
 {
 	struct vmap_area *va;
 	unsigned long free_area_size;
@@ -2749,5 +2749,23 @@ void get_vmalloc_info(struct vmalloc_info *vmi)
 out:
 	rcu_read_unlock();
 }
-#endif
 
+void get_vmalloc_info(struct vmalloc_info *vmi)
+{
+	static struct vmalloc_info cached_info;
+
+	if (!vmap_info_changed) {
+		*vmi = cached_info;
+		return;
+	}
+
+	WRITE_ONCE(vmap_info_changed, 0);
+	barrier();
+
+	calc_vmalloc_info(vmi);
+
+	barrier();
+	cached_info = *vmi;
+}
+
+#endif
-- 
2.1.4